v2.24.0-rc2
Now you can chat with PDFs directly in your RAG pipelines without extra steps.
Deep Dive
Haystack's v2.24.0-rc2 release introduces major upgrades for document processing. The new MarkdownHeaderSplitter intelligently chunks documents at headers while preserving hierarchy. Most importantly, developers can now pass PDFs and other files directly to OpenAI and Azure OpenAI chat generators using the new FileContent dataclass, eliminating complex preprocessing. The update also fixes pipeline mutation bugs and improves agent tool grouping, making complex RAG workflows significantly simpler to implement.
Why It Matters
This dramatically reduces the complexity of building production-ready RAG systems that need to handle real-world documents like research papers and reports.